{"total":156,"rows":[{"id":"2404.14082","title":"Mechanistic Interpretability for AI Safety -- A Review","created":"2024-04-22","cited_by_count":422,"area":"mechinterp"},{"id":"2501.17148","title":"AxBench: Steering LLMs? Even Simple Baselines Outperform Sparse Autoencoders","created":"2025-01-28","cited_by_count":218,"area":"mechinterp"},{"id":"2409.05907","title":"Programming Refusal with Conditional Activation Steering","created":"2024-09-06","cited_by_count":206,"area":"mechinterp"},{"id":"2404.16014","title":"Improving Dictionary Learning with Gated Sparse Autoencoders","created":"2024-04-24","cited_by_count":179,"area":"mechinterp"},{"id":"2409.14507","title":"A is for Absorption: Studying Feature Splitting and Absorption in Sparse Autoencoders","created":"2024-09-22","cited_by_count":160,"area":"mechinterp"},{"id":"2410.12877","title":"Improving Instruction-Following in Language Models through Activation Steering","created":"2024-10-15","cited_by_count":160,"area":"mechinterp"},{"id":"2407.12404","title":"Analyzing the Generalization and Reliability of Steering Vectors","created":"2024-07-17","cited_by_count":143,"area":"mechinterp"},{"id":"2411.14257","title":"Do I Know This Entity? Knowledge Awareness and Hallucinations in Language Models","created":"2024-11-21","cited_by_count":115,"area":"mechinterp"},{"id":"2410.13928","title":"Automatically Interpreting Millions of Features in Large Language Models","created":"2024-10-17","cited_by_count":113,"area":"mechinterp"},{"id":"2410.12299","title":"Semantics-Adaptive Activation Intervention for LLMs via Dynamic Steering Vectors","created":"2024-10-16","cited_by_count":63,"area":"mechinterp"},{"id":"2410.19750","title":"The Geometry of Concepts: Sparse Autoencoder Feature Structure","created":"2024-10-10","cited_by_count":52,"area":"mechinterp"},{"id":"2408.00113","title":"Measuring Progress in Dictionary Learning for Language Model Interpretability with Board Game Models","created":"2024-07-31","cited_by_count":51,"area":"mechinterp"},{"id":"2506.03093","title":"From Flat to Hierarchical: Extracting Sparse Representations with Matching Pursuit","created":"2025-06-03","cited_by_count":49,"area":"mechinterp"},{"id":"2410.14670","title":"Decomposing The Dark Matter of Sparse Autoencoders","created":"2024-10-18","cited_by_count":45,"area":"mechinterp"},{"id":"2410.08201","title":"Efficient Dictionary Learning with Switch Sparse Autoencoders","created":"2024-10-10","cited_by_count":39,"area":"mechinterp"},{"id":"2509.10625","title":"No Answer Needed: Predicting LLM Answer Accuracy from Question-Only Linear Probes","created":"2025-09-12","cited_by_count":38,"area":"mechinterp"},{"id":"2501.09929","title":"Interpretable Steering of Large Language Models with Feature Guided Activation Additions","created":"2025-01-17","cited_by_count":35,"area":"mechinterp"},{"id":"2411.04430","title":"Towards Unifying Interpretability and Control: Evaluation via Intervention","created":"2024-11-07","cited_by_count":32,"area":"mechinterp"},{"id":"2503.02080","title":"Linear Representations of Political Perspective Emerge in Large Language Models","created":"2025-03-03","cited_by_count":32,"area":"mechinterp"},{"id":"2604.28119","title":"Do Sparse Autoencoders Capture Concept Manifolds?","created":"2026-04-30","cited_by_count":31,"area":"mechinterp"},{"id":"2511.00617","title":"Belief Dynamics Reveal the Dual Nature of In-Context Learning and Activation Steering","created":"2025-11-01","cited_by_count":29,"area":"mechinterp"},{"id":"2410.06981","title":"Quantifying Feature Space Universality Across Large Language Models via Sparse Autoencoders","created":"2024-10-09","cited_by_count":27,"area":"mechinterp"},{"id":"2410.03334","title":"An X-Ray Is Worth 15 Features: Sparse Autoencoders for Interpretable Radiology Report Generation","created":"2024-10-04","cited_by_count":25,"area":"mechinterp"},{"id":"2505.20809","title":"Improved Representation Steering for Language Models","created":"2025-05-27","cited_by_count":21,"area":"mechinterp"},{"id":"2506.06609","title":"Transferring Linear Features Across Language Models With Model Stitching","created":"2025-06-07","cited_by_count":21,"area":"mechinterp"},{"id":"2605.11887","title":"Qwen-Scope: Turning Sparse Features into Development Tools for Large Language Models","created":"2026-05-12","cited_by_count":21,"area":"mechinterp"},{"id":"2409.04185","title":"Residual Stream Analysis with Multi-Layer SAEs","created":"2024-09-06","cited_by_count":19,"area":"mechinterp"},{"id":"2506.15679","title":"Dense SAE Latents Are Features, Not Bugs","created":"2025-06-18","cited_by_count":17,"area":"mechinterp"},{"id":"2412.17626","title":"Tracking the Feature Dynamics in LLM Training: A Mechanistic Study","created":"2024-12-23","cited_by_count":16,"area":"mechinterp"},{"id":"2506.01197","title":"Incorporating Hierarchical Semantics in Sparse Autoencoder Architectures","created":"2025-06-01","cited_by_count":14,"area":"mechinterp"},{"id":"2508.16560","title":"Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders","created":"2025-08-22","cited_by_count":14,"area":"mechinterp"},{"id":"2606.00995","title":"Subliminal Learning Is Steering Vector Distillation","created":"2026-05-31","cited_by_count":13,"area":"mechinterp"},{"id":"2508.12535","title":"CorrSteer: Generation-Time LLM Steering via Correlated Sparse Autoencoder Features","created":"2025-08-18","cited_by_count":12,"area":"mechinterp"},{"id":"2603.09313","title":"Curveball Steering: The Right Direction To Steer Isn't Always Linear","created":"2026-03-10","cited_by_count":11,"area":"mechinterp"},{"id":"2503.06269","title":"Using Mechanistic Interpretability to Craft Adversarial Attacks against Large Language Models","created":"2025-03-08","cited_by_count":10,"area":"mechinterp"},{"id":"2603.09972","title":"From Data Statistics to Feature Geometry: How Correlations Shape Superposition","created":"2026-03-10","cited_by_count":10,"area":"mechinterp"},{"id":"2602.03420","title":"CoCoEmo: Composable and Controllable Human-Like Emotional TTS via Activation Steering","created":"2026-02-03","cited_by_count":9,"area":"mechinterp"},{"id":"2503.24277","title":"Evaluating and Designing Sparse Autoencoders by Approximating Quasi-Orthogonality","created":"2025-03-31","cited_by_count":8,"area":"mechinterp"},{"id":"2508.01916","title":"Decomposing Representation Space into Interpretable Subspaces with Unsupervised Learning","created":"2025-08-03","cited_by_count":8,"area":"mechinterp"},{"id":"2509.22033","title":"OrtSAE: Orthogonal Sparse Autoencoders Uncover Atomic Features","created":"2025-09-26","cited_by_count":8,"area":"mechinterp"},{"id":"2602.06801","title":"On the Non-Identifiability of Steering Vectors in Large Language Models","created":"2026-02-06","cited_by_count":8,"area":"mechinterp"},{"id":"2603.06495","title":"COLD-Steer: Steering Large Language Models via In-Context One-step Learning Dynamics","created":"2026-03-06","cited_by_count":8,"area":"mechinterp"},{"id":"2501.19406","title":"Low-Rank Adapting Models for Sparse Autoencoders","created":"2025-01-31","cited_by_count":7,"area":"mechinterp"},{"id":"2512.13568","title":"Superposition as Lossy Compression: Measure with Sparse Autoencoders and Connect to Adversarial Vulnerability","created":"2025-12-15","cited_by_count":7,"area":"mechinterp"},{"id":"2506.18852","title":"Mechanistic Interpretability Needs Philosophy","created":"2025-06-23","cited_by_count":6,"area":"mechinterp"},{"id":"2508.16929","title":"Dimensional Collapse in Transformer Attention Outputs: A Challenge for Sparse Dictionary Learning","created":"2025-08-23","cited_by_count":6,"area":"mechinterp"},{"id":"2509.23323","title":"LLM Interpretability with Identifiable Temporal-Instantaneous Representation","created":"2025-09-27","cited_by_count":6,"area":"mechinterp"},{"id":"2511.19265","title":"Unboxing the Black Box: Mechanistic Interpretability for Algorithmic Understanding of Neural Networks","created":"2025-11-24","cited_by_count":6,"area":"mechinterp"},{"id":"2602.11881","title":"From Atoms to Trees: Building a Structured Feature Forest with Hierarchical Sparse Autoencoders","created":"2026-02-12","cited_by_count":6,"area":"mechinterp"},{"id":"2603.02237","title":"Concept Heterogeneity-aware Representation Steering","created":"2026-02-13","cited_by_count":6,"area":"mechinterp"},{"id":"2603.28744","title":"Stop Probing, Start Coding: Why Linear Probes and Sparse Autoencoders Fail at Compositional Generalisation","created":"2026-03-30","cited_by_count":6,"area":"mechinterp"},{"id":"2604.02608","title":"Steerable but Not Decodable: Function Vectors Operate Beyond the Logit Lens","created":"2026-04-03","cited_by_count":6,"area":"mechinterp"},{"id":"2605.05892","title":"Beyond Steering Vector: Flow-based Activation Steering for Inference-Time Intervention","created":"2026-05-07","cited_by_count":6,"area":"mechinterp"},{"id":"2607.07003","title":"Dissociating the Internal Representations of Sycophancy in LLMs","created":"2026-07-08","cited_by_count":6,"area":"mechinterp"},{"id":"2509.16820","title":"DISCO: Disentangled Communication Steering for Large Language Models","created":"2025-09-20","cited_by_count":5,"area":"mechinterp"},{"id":"2510.02917","title":"Mechanistic Interpretability of Code Correctness in LLMs via Sparse Autoencoders","created":"2025-10-03","cited_by_count":5,"area":"mechinterp"},{"id":"2510.16820","title":"Finding Manifolds With Bilinear Autoencoders","created":"2025-10-19","cited_by_count":5,"area":"mechinterp"},{"id":"2603.12298","title":"Global Evolutionary Steering: Refining Activation Steering Control via Cross-Layer Consistency","created":"2026-03-12","cited_by_count":5,"area":"mechinterp"},{"id":"2603.13249","title":"Steering at the Source: Style Modulation Heads for Robust Persona Control","created":"2026-02-24","cited_by_count":5,"area":"mechinterp"},{"id":"2605.01167","title":"Minimizing Collateral Damage in Activation Steering","created":"2026-05-01","cited_by_count":5,"area":"mechinterp"},{"id":"2605.03907","title":"Steer Like the LLM: Activation Steering that Mimics Prompting","created":"2026-05-05","cited_by_count":5,"area":"mechinterp"},{"id":"2402.03855","title":"Challenges in Mechanistically Interpreting Model Representations","created":"2024-02-06","cited_by_count":4,"area":"mechinterp"},{"id":"2503.11429","title":"Combining Causal Models for More Accurate Abstractions of Neural Networks","created":"2025-03-14","cited_by_count":4,"area":"mechinterp"},{"id":"2511.20798","title":"Physics Steering: Causal Control of Cross-Domain Concepts in a Physics Foundation Model","created":"2025-11-25","cited_by_count":4,"area":"mechinterp"},{"id":"2604.15557","title":"Predicting Where Steering Vectors Succeed","created":"2026-04-16","cited_by_count":4,"area":"mechinterp"},{"id":"2605.07922","title":"Tree SAE: Learning Hierarchical Feature Structures in Sparse Autoencoders","created":"2026-05-08","cited_by_count":4,"area":"mechinterp"},{"id":"2606.06333","title":"Subspace-Aware Sparse Autoencoders for Effective Mechanistic Interpretability","created":"2026-06-04","cited_by_count":4,"area":"mechinterp"},{"id":"2412.07947","title":"GPT-2 Through the Lens of Vector Symbolic Architectures","created":"2024-12-10","cited_by_count":3,"area":"mechinterp"},{"id":"2509.03738","title":"Mechanistic Interpretability with Sparse Autoencoder Neural Operators","created":"2025-09-03","cited_by_count":3,"area":"mechinterp"},{"id":"2509.25045","title":"Hyperdimensional Probe: Decoding LLM Representations via Vector Symbolic Architectures","created":"2025-09-29","cited_by_count":3,"area":"mechinterp"},{"id":"2510.27413","title":"Atlas-Alignment: Making Interpretability Transferable Across Language Models","created":"2025-10-31","cited_by_count":3,"area":"mechinterp"},{"id":"2512.05534","title":"A Unified Theory of Sparse Dictionary Learning in Mechanistic Interpretability: Piecewise Biconvexity and Spurious Minima","created":"2025-12-05","cited_by_count":3,"area":"mechinterp"},{"id":"2512.07355","title":"A Geometric Unification of Concept Learning with Concept Cones","created":"2025-12-08","cited_by_count":3,"area":"mechinterp"},{"id":"2605.06342","title":"Don't Lose Focus: Activation Steering via Key-Orthogonal Projections","created":"2026-05-07","cited_by_count":3,"area":"mechinterp"},{"id":"2605.14694","title":"The Rate-Distortion-Polysemanticity Tradeoff in SAEs","created":"2026-05-14","cited_by_count":3,"area":"mechinterp"},{"id":"2605.31183","title":"Steering LLMs? Actually, Sparse Autoencoders can outperform simple baselines","created":"2026-05-29","cited_by_count":3,"area":"mechinterp"},{"id":"2606.04775","title":"Activation Steering of Video Generation Models via Reduced-Order Linear Optimal Control","created":"2026-06-03","cited_by_count":3,"area":"mechinterp"},{"id":"2506.01247","title":"Beyond Interpretability: When, Why, and How Sparse Autoencoders Enable Label-Free Visual Steering","created":"2025-06-02","cited_by_count":2,"area":"mechinterp"},{"id":"2506.07406","title":"InverseScope: Scalable Activation Inversion for Interpreting Large Language Models","created":"2025-06-09","cited_by_count":2,"area":"mechinterp"},{"id":"2506.18598","title":"No Training Wheels: Steering Vectors for Bias Correction at Inference Time","created":"2025-06-23","cited_by_count":2,"area":"mechinterp"},{"id":"2508.16989","title":"Unveiling the Latent Directions of Reflection in Large Language Models","created":"2025-08-23","cited_by_count":2,"area":"mechinterp"},{"id":"2510.26411","title":"MedSAE: Dissecting MedCLIP Representations with Sparse Autoencoders","created":"2025-10-30","cited_by_count":2,"area":"mechinterp"},{"id":"2601.20028","title":"Decomposing multimodal embedding spaces with group-sparse autoencoders","created":"2026-01-27","cited_by_count":2,"area":"mechinterp"},{"id":"2602.04718","title":"Superposition Without Interference? Towards Isolated Interventions via Almost Orthogonal Features in Language Models","created":"2026-02-04","cited_by_count":2,"area":"mechinterp"},{"id":"2606.08682","title":"Activation Steering Induces Emergent Misalignment: A More Comprehensive Evaluation","created":"2026-06-07","cited_by_count":2,"area":"mechinterp"},{"id":"2606.09607","title":"Closure-Validated Circuit Discovery in Attention Heads: Co-activation Proposes, Ablation Disposes","created":"2026-06-08","cited_by_count":2,"area":"mechinterp"},{"id":"2606.11205","title":"Dual-Stance Evaluation of Sycophancy: The Structure of Agreement and the Limits of Intervention","created":"2026-04-22","cited_by_count":2,"area":"mechinterp"},{"id":"2607.10517","title":"Conditional Optimal Bridge for Riemannian Activation Steering","created":"2026-07-12","cited_by_count":2,"area":"mechinterp"},{"id":"2411.18247","title":"A gentle push funziona benissimo: making instructed models in Italian via contrastive activation steering","created":"2024-11-27","cited_by_count":1,"area":"mechinterp"},{"id":"2502.03618","title":"The Logical Implication Steering Method for Conditional Interventions on Transformer Generation","created":"2025-02-05","cited_by_count":1,"area":"mechinterp"},{"id":"2506.05239","title":"Evaluating Sparse Autoencoders: From Shallow Design to Matching Pursuit","created":"2025-06-05","cited_by_count":1,"area":"mechinterp"},{"id":"2512.15987","title":"Provably Extracting the Features from a General Superposition","created":"2025-12-17","cited_by_count":1,"area":"mechinterp"},{"id":"2602.09784","title":"Circuit Fingerprints: How Answer Tokens Encode Their Geometrical Path","created":"2026-02-10","cited_by_count":1,"area":"mechinterp"},{"id":"2602.13524","title":"Singular Vectors of Attention Heads Align with Features","created":"2026-02-13","cited_by_count":1,"area":"mechinterp"},{"id":"2603.13259","title":"How Transformers Reject Wrong Answers: Rotational Dynamics of Factual Constraint Processing","created":"2026-02-25","cited_by_count":1,"area":"mechinterp"},{"id":"2604.24936","title":"A Unifying Framework for Unsupervised Concept Extraction","created":"2026-04-27","cited_by_count":1,"area":"mechinterp"},{"id":"2605.08891","title":"Bilinear autoencoders find interpretable manifolds","created":"2026-05-09","cited_by_count":1,"area":"mechinterp"},{"id":"2605.09224","title":"SMIXAE: Towards Unsupervised Manifold Discovery in Language Models","created":"2026-05-09","cited_by_count":1,"area":"mechinterp"},{"id":"2605.09887","title":"The Geometric Wall: Manifold Structure Predicts Layerwise Sparse Autoencoder Scaling Laws","created":"2026-05-11","cited_by_count":1,"area":"mechinterp"},{"id":"2605.09967","title":"Tensor Product Representation Probes Reveal Shared Structure Across Linear Directions","created":"2026-05-11","cited_by_count":1,"area":"mechinterp"}],"next_cursor":"100"}